The exoskeleton analogy seems to be fitting where my work-mode is configurable: moving from tentative to trusting. But the AI needs to be explicitly set up to learn my every action. Currently this is a chore at best, just impossible in other cases.
Hmm, may be you're right about UX designers not wanting to explain. For instance, I certainly won't want that vertical bar for aesthetic reasons. It's just hard to defend objectively.
They're just advertising the tech and the platform. Probably so people could come up with some real use cases and apps for it... otherwise what's the point of introducing it a year earlier.